He tested every IP KVM in his homelab — one of them caught fire

Jeff Geerling ran the cheap remote-control boxes that let you reach a headless machine's BIOS over the network. The thread out-tested the test: a YC founder wire-debugged the USB signal, and one commenter's NanoKVM burst into flames 15 minutes after plugging in.

Why a homelab staple suddenly needed a shoot-out

IP KVMs — little devices that emulate a USB keyboard, mouse, and monitor so you can reach a frozen or pre-boot machine from anywhere — have gone from niche to crowded, with a dozen sub-$100 options on Amazon. Jeff Geerling tested the field across his homelab. The blog returned a 403 to our fetch, so the device-by-device specifics below are drawn from his conclusions as discussed in the thread and the comparison the commenters built around them.

The field, and where each one landed

The rough verdict that the thread coalesced around:

Device Rough price Read from the discussion
PiKVM V4 Plus ~$250+ The do-it-right pick; correct USB behavior, Redfish support, not cheap
JetKVM ~$69 Crowd favorite for the price; great Tailscale support, HDMI adapter quirks
GL.iNet (Comet line) ~$70+ Nicer UI, but harder to customize; one USB edge-case bug
Sipeed NanoKVM ~$50 Cheapest; one unit caught fire, and a security scare
TinyPilot ~$300 "More polished feel" than DIY PiKVM
DIY PiKVM (Pi + parts) ~$100 less Fully supported in their docs; messy software install

The single most useful contribution came not from the post but from a founder who'd run these in production. His team refurbishes laptops with robots, which means a KVM has to send flawless BIOS keystrokes:

"We were a GL.iNet KVM shop until we ran into a nasty issue with a specific ThinkPad - the GL.iNet would send an incorrect USB 0 byte which most laptops ignored, except this ThinkPad… I got a low level USB debugger and wire-debugged the USB signal… The PiKVM was doing things properly, so we swapped all (~10) of our KVMs for it." — a YC startup co-founder who wire-debugged the USB signal

That is the kind of finding the post's critics said it lacked. The harshest review of the review:

"It seems like basically a listicle with 'this is great' and a copy/paste of the datasheet specs… I was expecting a list of facets that were tested, with insight you can only get from having the hardware in front of you." — a reader who wanted real reboot/BIOS-access testing

The cheap end is where it gets dangerous

The budget tier came with literal smoke. One commenter's account of the Sipeed NanoKVM is the cautionary tale of the thread:

"I bought a SiPeed NanoKVM. It caught fire 15 minutes after being plugged in. Despite providing pictures of the charred PCB, they insisted I ship it back… Maybe just buy the JetKVM. It looks nice!" — a homelabber whose NanoKVM caught fire

The security footnote is worse than aesthetics: a commenter linked reporting on disclosed NanoKVM vulnerabilities, prompting the half-joking question of whether ordering one risks "a visit from the FBI." Meanwhile veterans pointed past the whole category — one noted Intel vPro AMT is "basically an always-on KVM that's part of CPU firmware," already sitting in many machines.

What the test was actually for

Strip away the listicle complaint and a real buyer's map emerges: pay up for PiKVM if correctness and BIOS reliability matter, grab a JetKVM if you want cheap and pleasant and can live with an HDMI adapter, and treat the $50 tier as a fire-and-vulnerability gamble. The most valuable testing in this round of "I tested every IP KVM" happened in the comments, where someone with a USB protocol analyzer proved which box sends correct keystrokes — and which one a single ThinkPad could tell was lying.

Get breakdowns like this every day

The top Hacker News stories, unpacked — by email, Telegram, or RSS.

Next story · 4 of 25 Read more
🌍Other

An essay says interviews are random functions — then tptacek says its own fix misses the point

The argument is familiar: technical interviews filter for the wrong things and reject good engineers. The twist is in the thread, where the highest-karma commenter on the site dismantles the essay's proposed metric for not measuring the one thing that matters.

★ 54 · 💬 102 · fagnerbrack.com · 5 min